Critical Remarks on the Validity of Fatigue Life Evaluation Methods Based on Local Stress-Strain Behavior

Abstract

The conventional fatigue analysis using nominal stresses and Miner's rule is known to give inaccurate life predictions. It has been speculated that the major deficiencies can be overcome if local stresses and strains in the immediate vicinity of a stress raiser, instead of nominal stresses, are considered in the estimation of fatigue life. At present, there are a number of proposals which consider local stress-strain concepts, some of which are already developed to a state of practical applicability. In this paper, the best-known concepts are reviewed. It was found that all methods involving the determination of the stress-strain behavior at the notch root retained the conventional methods for computing damage accumulation. The question which arises is: What improvement in life prediction can be expected eventually when the correct local stress-strain history is known, but the damage due to local stress-strain cycles is linearly accumulated? Open-hole specimens of the aluminum alloys 2024-T3 and 7075-T6 were subjected to a flight-by-flight load program. The corresponding local strain and stress histories were determined by local strain measurements and the companion specimen method, respectively. Fatigue life was estimated by three different methods using: (a) the local stress cycles and unnotched specimen applied stress-cycles to failure (S-N) data, (b) the nominal stress cycles and S-N data from prestressed open-hole specimens, and (c) for comparison, the nominal stress cycles and S-N data from the open-hole specimen. In each case the damage was linearly accumulated. For the flight-by-flight testing sequences, the companion specimen method indicated that the residual stresses at the notch root stabilized asymptotically to a limiting value at between 5 and 20 percent of the life-to-crack initiation. Virgin open-hole specimens were also prestressed to develop, at the hole, the stress-strain status of the stabilized residual stresses determined previously during flight-by-flight loading. The prestressed specimens were subjected to constant amplitude loading and the resulting lives used for the damage calculation in Method B. The ratios of test life to estimated life obtained using the three methods were compared and the following conclusions drawn: with increasing accuracy of the incorporation of local stress-strain behavior in future analysis methods (and using linear damage accumulation), the predicted life may still differ from the actual life. The estimated lives may be both conservative and unconservative, so that a scatter of life estimates of 1:3.0 is possible. This is, however, smaller than that of the life estimates using the conventional nominal stress concept which was found to be 1:6.0.